agora inbox for pgsql-hackers@postgresql.orghelp / color / mirror / Atom feed
[PATCH v44 07/10] Split cluster.h to create repack_internal.h 354+ messages / 1 participants [nested] [flat]
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v44 07/10] Split cluster.h to create repack_internal.h @ 2026-03-24 14:09 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-03-24 14:09 UTC (permalink / raw) Most of the details of concurrent repack are only needed by pgoutput_repack; and they have a nasty effect on headers included by cluster.h (used by several high-profile places), so isolate that for cleanliness. Also, change_useless_for_repack() is better declared in decode.h. --- src/backend/commands/cluster.c | 128 ++++++++++-------- src/backend/postmaster/bgworker.c | 2 +- .../pgoutput_repack/pgoutput_repack.c | 13 +- src/include/commands/cluster.h | 57 +------- src/include/commands/repack_internal.h | 59 ++++++++ src/include/replication/decode.h | 4 + 6 files changed, 147 insertions(+), 116 deletions(-) create mode 100644 src/include/commands/repack_internal.h diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c index 75556cbdafb..2c3058ba10d 100644 --- a/src/backend/commands/cluster.c +++ b/src/backend/commands/cluster.c @@ -59,6 +59,7 @@ #include "commands/cluster.h" #include "commands/defrem.h" #include "commands/progress.h" +#include "commands/repack_internal.h" #include "commands/tablecmds.h" #include "commands/vacuum.h" #include "executor/executor.h" @@ -205,17 +206,11 @@ typedef struct DecodingWorkerShared char error_queue[FLEXIBLE_ARRAY_MEMBER]; } DecodingWorkerShared; -/* - * Generate worker's output file name. If relations of the same 'relid' happen - * to be processed at the same time, they must be from different databases and - * therefore different backends must be involved. (PID is already present in - * the fileset name.) - */ -static inline void -DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) -{ - snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); -} +/* Is this process a REPACK worker? */ +static bool is_repack_worker = false; + +static pid_t backend_pid; +static ProcNumber backend_proc_number; /* * Backend-local information to control the decoding worker. @@ -241,6 +236,7 @@ static DecodingWorker *decoding_worker = NULL; */ volatile sig_atomic_t RepackMessagePending = false; +static LOCKMODE RepackLockLevel(bool concurrent); static bool cluster_rel_recheck(RepackCommand cmd, Relation OldHeap, Oid indexOid, Oid userid, LOCKMODE lmode, int options); @@ -298,12 +294,16 @@ static Relation process_single_relation(RepackStmt *stmt, ClusterParams *params); static Oid determine_clustered_index(Relation rel, bool usingindex, const char *indexname); -static void start_decoding_worker(Oid relid); -static void stop_decoding_worker(void); + +static void start_repack_decoding_worker(Oid relid); +static void stop_repack_decoding_worker(void); static void repack_worker_internal(dsm_segment *seg); static void export_initial_snapshot(Snapshot snapshot, DecodingWorkerShared *shared); static Snapshot get_initial_snapshot(DecodingWorker *worker); +extern bool am_decoding_for_repack(void); +static void DecodingWorkerFileName(char *fname, Oid relid, uint32 seq); + static void ProcessRepackMessage(StringInfo msg); static const char *RepackCommandAsString(RepackCommand cmd); @@ -369,17 +369,8 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) parser_errposition(pstate, opt->location)); } - /* - * Determine the lock mode expected by cluster_rel(). - * - * In the exclusive case, we obtain AccessExclusiveLock right away to - * avoid lock-upgrade hazard in the single-transaction case. In the - * CONCURRENTLY case, the AccessExclusiveLock will only be used at the end - * of processing, supposedly for very short time. Until then, we'll have - * to unlock the relation temporarily, so there's no lock-upgrade hazard. - */ - lockmode = (params.options & CLUOPT_CONCURRENT) == 0 ? - AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lockmode = RepackLockLevel((params.options & CLUOPT_CONCURRENT) != 0); /* * If a single relation is specified, process it and we're done ... unless @@ -434,6 +425,12 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) "Repack", ALLOCSET_DEFAULT_SIZES); + /* + * Since we open a new transaction for each relation, we have to check + * that the relation still is what we think it is. + * + * In single-transaction CLUSTER, we don't need the overhead. + */ params.options |= CLUOPT_RECHECK; /* @@ -544,6 +541,22 @@ ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel) MemoryContextDelete(repack_context); } +/* + * In the non-concurrent case, we obtain AccessExclusiveLock throughout the + * operation to avoid any lock-upgrade hazards. In the concurrent case, we + * grab ShareUpdateExclusiveLock (jsut like VACUUM) for most of the + * processing and only acquire AccessExclusiveLock at the end, to swap the + * relation -- supposedly for a short time. + */ +static LOCKMODE +RepackLockLevel(bool concurrent) +{ + if (concurrent) + return ShareUpdateExclusiveLock; + else + return AccessExclusiveLock; +} + /* * cluster_rel * @@ -583,25 +596,22 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, bool concurrent = ((params->options & CLUOPT_CONCURRENT) != 0); Oid ident_idx = InvalidOid; - /* - * The lock mode is AccessExclusiveLock for normal processing and - * ShareUpdateExclusiveLock for concurrent processing (so that SELECT, - * INSERT, UPDATE and DELETE commands work, but cluster_rel() cannot be - * called concurrently for the same relation). - */ - lmode = !concurrent ? AccessExclusiveLock : ShareUpdateExclusiveLock; + /* Determine the lock mode to use. */ + lmode = RepackLockLevel(concurrent); - /* There are specific requirements on concurrent processing. */ + /* + * Check some preconditions in the concurrent case. This also obtains the + * replica index OID. + */ if (concurrent) { /* - * Make sure we have no XID assigned, otherwise call of - * repack_setup_logical_decoding() can cause a deadlock. + * Make sure we're not in a transaction block. * - * The existence of transaction block actually does not imply that XID - * was already assigned, but it very likely is. We might want to check - * the result of GetCurrentTransactionIdIfAny() instead, but that - * would be less clear from user's perspective. + * The reason is that repack_setup_logical_decoding() could deadlock + * if there's an XID already assigned. It would be possible to run in + * a transaction block if we had no XID, but this restriction is + * simpler for users to understand and we don't lose anything. */ PreventInTransactionBlock(isTopLevel, "REPACK (CONCURRENTLY)"); @@ -626,15 +636,11 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, RestrictSearchPath(); /* - * Since we may open a new transaction for each relation, we have to check - * that the relation still is what we think it is. + * Recheck that the relation is still what it was when we started. * - * If this is a single-transaction CLUSTER, we can skip these tests. We - * *must* skip the one on indisclustered since it would reject an attempt - * to cluster a not-previously-clustered index. - * - * XXX move [some of] these comments to where the RECHECK flag is - * determined? + * Note that it's critical to skip this in single-relation CLUSTER; + * otherwise, we would reject an attempt to cluster using a + * not-previously-clustered index. */ if (recheck && !cluster_rel_recheck(cmd, OldHeap, indexOid, save_userid, @@ -754,7 +760,7 @@ cluster_rel(RepackCommand cmd, Relation OldHeap, Oid indexOid, * However it still seems a good practice to make sure that the * worker never survives the REPACK command. */ - stop_decoding_worker(); + stop_repack_decoding_worker(); } } PG_END_TRY(); @@ -1093,7 +1099,7 @@ rebuild_relation(Relation OldHeap, Relation index, bool verbose, * clustering index) and checking again if it's still eligible for * REPACK CONCURRENTLY. */ - start_decoding_worker(tableOid); + start_repack_decoding_worker(tableOid); /* * Wait until the worker has the initial snapshot and retrieve it. @@ -2538,7 +2544,6 @@ RepackCommandAsString(RepackCommand cmd) return "???"; /* keep compiler quiet */ } - /* * Is this backend performing logical decoding on behalf of REPACK * (CONCURRENTLY) ? @@ -3688,7 +3693,7 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes) * contents to a new table. */ static void -start_decoding_worker(Oid relid) +start_repack_decoding_worker(Oid relid) { Size size; dsm_segment *seg; @@ -3779,7 +3784,7 @@ start_decoding_worker(Oid relid) * we need to stop it explicitly at least on ERROR in the launching backend. */ static void -stop_decoding_worker(void) +stop_repack_decoding_worker(void) { BgwHandleStatus status; @@ -3817,12 +3822,6 @@ stop_decoding_worker(void) decoding_worker = NULL; } -/* Is this process a REPACK worker? */ -static bool is_repack_worker = false; - -static pid_t backend_pid; -static ProcNumber backend_proc_number; - /* * See ParallelWorkerShutdown for details. */ @@ -3869,7 +3868,7 @@ RepackWorkerMain(Datum main_arg) /* * Join locking group - see the comments around the call of - * start_decoding_worker(). + * start_repack_decoding_worker(). */ if (!BecomeLockGroupMember(shared->backend_proc, backend_pid)) /* The leader is not running anymore. */ @@ -4039,6 +4038,18 @@ get_initial_snapshot(DecodingWorker *worker) return snapshot; } +/* + * Generate worker's output file name. If relations of the same 'relid' happen + * to be processed at the same time, they must be from different databases and + * therefore different backends must be involved. (PID is already present in + * the fileset name.) + */ +static void +DecodingWorkerFileName(char *fname, Oid relid, uint32 seq) +{ + snprintf(fname, MAXPGPATH, "%u-%u", relid, seq); +} + bool IsRepackWorker(void) { @@ -4067,7 +4078,6 @@ void ProcessRepackMessages(void) { MemoryContext oldcontext; - static MemoryContext hpm_context = NULL; /* diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index f8a8d1681e9..9e876d55e27 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/parallel.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "libpq/pqsignal.h" #include "miscadmin.h" #include "pgstat.h" diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c index 032fbd0e5b0..cc9ce615b18 100644 --- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c +++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c @@ -13,7 +13,7 @@ #include "postgres.h" #include "access/detoast.h" -#include "commands/cluster.h" +#include "commands/repack_internal.h" #include "replication/snapbuild.h" #include "utils/memutils.h" @@ -176,7 +176,7 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, ConcurrentChangeKind kind, HeapTuple tuple) { RepackDecodingState *dstate; - MemoryContext oldcxt; + MemoryContext oldcxt; BufFile *file; List *attrs_ext = NIL; int natt_ext; @@ -226,7 +226,10 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, slot_getsomeattrs(slot, i + 1); - /* This is a non-null varlena datum, but we only care if it's out-of-line */ + /* + * This is a non-null varlena datum, but we only care if it's + * out-of-line + */ varlen = (varlena *) DatumGetPointer(slot->tts_values[i]); if (!VARATT_IS_EXTERNAL(varlen)) continue; @@ -244,8 +247,8 @@ repack_store_change(LogicalDecodingContext *ctx, Relation relation, * attributes (those actually should never appear on disk), so * only TOASTed attribute can be seen here. * - * FIXME in what circumstances can an ONDISK attr appear? - * Why aren't these written separately? + * FIXME in what circumstances can an ONDISK attr appear? Why + * aren't these written separately? */ Assert(VARATT_IS_EXTERNAL_ONDISK(varlen)); } diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h index 1c0ac3ab4f5..1528d34fa42 100644 --- a/src/include/commands/cluster.h +++ b/src/include/commands/cluster.h @@ -13,17 +13,12 @@ #ifndef CLUSTER_H #define CLUSTER_H -#include "nodes/execnodes.h" +#include <signal.h> + #include "nodes/parsenodes.h" #include "parser/parse_node.h" -#include "replication/decode.h" -#include "postmaster/bgworker.h" -#include "replication/logical.h" -#include "storage/buffile.h" #include "storage/lockdefs.h" -#include "storage/shm_mq.h" #include "utils/relcache.h" -#include "utils/resowner.h" /* flag bits for ClusterParams->options */ @@ -34,55 +29,14 @@ #define CLUOPT_ANALYZE 0x08 /* do an ANALYZE */ #define CLUOPT_CONCURRENT 0x10 /* allow concurrent data changes */ - /* options for CLUSTER */ typedef struct ClusterParams { bits32 options; /* bitmask of CLUOPT_* */ } ClusterParams; - -/* - * The following definitions are used by REPACK CONCURRENTLY. - */ - -/* - * Stored as a single byte in the output file. - */ -#define CHANGE_INSERT 'i' -#define CHANGE_UPDATE_OLD 'u' -#define CHANGE_UPDATE_NEW 'U' -#define CHANGE_DELETE 'd' -typedef char ConcurrentChangeKind; - -/* - * Logical decoding state. - * - * The output plugin uses it to store the data changes that it decodes from - * WAL while the table contents is being copied to a new storage. - */ -typedef struct RepackDecodingState -{ -#ifdef USE_ASSERT_CHECKING - /* The relation whose changes we're decoding. */ - Oid relid; -#endif - - /* Per-change memory context. */ - MemoryContext change_cxt; - - /* A tuple slot used to pass tuples back and forth */ - TupleTableSlot *slot; - - /* The current output file. */ - BufFile *file; -} RepackDecodingState; - extern PGDLLIMPORT volatile sig_atomic_t RepackMessagePending; -extern bool IsRepackWorker(void); -extern void HandleRepackMessageInterrupt(void); -extern void ProcessRepackMessages(void); extern void ExecRepack(ParseState *pstate, RepackStmt *stmt, bool isTopLevel); @@ -104,8 +58,9 @@ extern void finish_heap_swap(Oid OIDOldHeap, Oid OIDNewHeap, MultiXactId cutoffMulti, char newrelpersistence); -extern bool am_decoding_for_repack(void); -extern bool change_useless_for_repack(XLogRecordBuffer *buf); +extern bool IsRepackWorker(void); +extern void HandleRepackMessageInterrupt(void); +extern void ProcessRepackMessages(void); + -extern void RepackWorkerMain(Datum main_arg); #endif /* CLUSTER_H */ diff --git a/src/include/commands/repack_internal.h b/src/include/commands/repack_internal.h new file mode 100644 index 00000000000..f90c973f5a0 --- /dev/null +++ b/src/include/commands/repack_internal.h @@ -0,0 +1,59 @@ +/*------------------------------------------------------------------------- + * + * repack_internal.h + * header for REPACK internals + * + * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group + * Portions Copyright (c) 1994-5, Regents of the University of California + * + * src/include/commands/repack_internal.h + * + *------------------------------------------------------------------------- + */ +#ifndef REPACK_INTERNAL_H +#define REPACK_INTERNAL_H + +#include "nodes/execnodes.h" +#include "replication/decode.h" +#include "postmaster/bgworker.h" +#include "replication/logical.h" +#include "storage/buffile.h" +#include "storage/shm_mq.h" +#include "utils/resowner.h" + +/* + * Stored as a single byte in the output file. + */ +#define CHANGE_INSERT 'i' +#define CHANGE_UPDATE_OLD 'u' +#define CHANGE_UPDATE_NEW 'U' +#define CHANGE_DELETE 'd' +typedef char ConcurrentChangeKind; + +/* + * Logical decoding state. + * + * The output plugin uses it to store the data changes that it decodes from + * WAL while the table contents is being copied to a new storage. + */ +typedef struct RepackDecodingState +{ +#ifdef USE_ASSERT_CHECKING + /* The relation whose changes we're decoding. */ + Oid relid; +#endif + + /* Per-change memory context. */ + MemoryContext change_cxt; + + /* A tuple slot used to pass tuples back and forth */ + TupleTableSlot *slot; + + /* The current output file. */ + BufFile *file; +} RepackDecodingState; + + +extern void RepackWorkerMain(Datum main_arg); + +#endif /* REPACK_INTERNAL_H */ diff --git a/src/include/replication/decode.h b/src/include/replication/decode.h index 49f00fc48b8..02b5393474c 100644 --- a/src/include/replication/decode.h +++ b/src/include/replication/decode.h @@ -31,4 +31,8 @@ extern void logicalmsg_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf extern void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *record); +/* in commands/cluster.c */ +extern bool change_useless_for_repack(XLogRecordBuffer *buf); + + #endif -- 2.47.3 --gwom7bl7ogtszo4k Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="v44-0008-Use-BulkInsertState-when-copying-data-to-the-new.patch" ^ permalink raw reply [nested|flat] 354+ messages in thread
* [PATCH v54 5/5] Reserve replication slots specifically for REPACK @ 2026-04-01 17:54 Álvaro Herrera <alvherre@kurilemu.de> 0 siblings, 0 replies; 354+ messages in thread From: Álvaro Herrera @ 2026-04-01 17:54 UTC (permalink / raw) This allows REPACK to not interfere with other operations that use replication slots. This eases configurability. --- doc/src/sgml/config.sgml | 16 ++++ doc/src/sgml/ref/repack.sgml | 6 +- src/backend/commands/repack.c | 2 +- src/backend/commands/repack_worker.c | 7 +- src/backend/replication/logical/launcher.c | 2 +- src/backend/replication/logical/logical.c | 8 +- .../replication/logical/logicalfuncs.c | 2 +- src/backend/replication/logical/slotsync.c | 5 +- src/backend/replication/slot.c | 86 ++++++++++++------- src/backend/replication/slotfuncs.c | 19 ++-- src/backend/replication/walsender.c | 9 +- src/backend/utils/misc/guc_parameters.dat | 8 ++ src/backend/utils/misc/postgresql.conf.sample | 2 + src/include/replication/logical.h | 3 +- src/include/replication/slot.h | 5 +- 15 files changed, 117 insertions(+), 63 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index d3fea738ca3..a90d163e416 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -4638,6 +4638,22 @@ restore_command = 'copy "C:\\server\\archivedir\\%f" "%p"' # Windows </listitem> </varlistentry> + <varlistentry id="guc-max-repack-replication-slots" xreflabel="max_repack_replication_slots"> + <term><varname>max_repack_replication_slots</varname> (<type>integer</type>) + <indexterm> + <primary><varname>max_repack_replication_slots</varname> configuration parameter</primary> + </indexterm> + </term> + <listitem> + <para> + Specifies the maximum number of replication slots for use of + the <command>REPACK</command> command. The default is 5. + This parameter can only be set at server start. + </para> + </listitem> + </varlistentry> + + <varlistentry id="guc-max-replication-slots" xreflabel="max_replication_slots"> <term><varname>max_replication_slots</varname> (<type>integer</type>) <indexterm> diff --git a/doc/src/sgml/ref/repack.sgml b/doc/src/sgml/ref/repack.sgml index e993dfb3108..c532a39ee07 100644 --- a/doc/src/sgml/ref/repack.sgml +++ b/doc/src/sgml/ref/repack.sgml @@ -293,9 +293,9 @@ REPACK [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] USING <listitem> <para> - The <link linkend="guc-max-replication-slots"><varname>max_replication_slots</varname></link> - configuration parameter does not allow for creation of an additional - replication slot. + The <link linkend="guc-max-repack-replication-slots"><varname>max_repack_replication_slots</varname></link> + configuration parameter does not allow for the creation of an + additional replication slot. </para> </listitem> </itemizedlist> diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c index d1be6c60e88..ec3a2cd441d 100644 --- a/src/backend/commands/repack.c +++ b/src/backend/commands/repack.c @@ -3382,7 +3382,7 @@ start_repack_decoding_worker(Oid relid) ereport(ERROR, errcode(ERRCODE_CONFIGURATION_LIMIT_EXCEEDED), errmsg("out of background worker slots"), - errhint("You might need to increase \"%s\".", "max_worker_processes")); + errhint("You might need to increase \"%s\".", "max_repack_replication_slots")); decoding_worker->seg = seg; decoding_worker->error_mqh = mqh; diff --git a/src/backend/commands/repack_worker.c b/src/backend/commands/repack_worker.c index c85166ba849..39cbb0252f4 100644 --- a/src/backend/commands/repack_worker.c +++ b/src/backend/commands/repack_worker.c @@ -223,7 +223,7 @@ repack_setup_logical_decoding(Oid relid) * Make sure we can use logical decoding. */ CheckSlotPermissions(); - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(true); /* * A single backend should not execute multiple REPACK commands at a time, @@ -232,8 +232,8 @@ repack_setup_logical_decoding(Oid relid) * RS_TEMPORARY so that the slot gets cleaned up on ERROR. */ snprintf(NameStr(slotname), NAMEDATALEN, "repack_%d", MyProcPid); - ReplicationSlotCreate(NameStr(slotname), true, RS_TEMPORARY, false, false, - false); + ReplicationSlotCreate(NameStr(slotname), true, RS_TEMPORARY, false, true, + false, false); EnsureLogicalDecodingEnabled(); @@ -244,6 +244,7 @@ repack_setup_logical_decoding(Oid relid) ctx = CreateInitDecodingContext(REPL_PLUGIN_NAME, NIL, true, + true, InvalidXLogRecPtr, XL_ROUTINE(.page_read = read_local_xlog_page, .segment_open = wal_segment_open, diff --git a/src/backend/replication/logical/launcher.c b/src/backend/replication/logical/launcher.c index 09964198550..d83125afd0d 100644 --- a/src/backend/replication/logical/launcher.c +++ b/src/backend/replication/logical/launcher.c @@ -1575,7 +1575,7 @@ CreateConflictDetectionSlot(void) errmsg("creating replication conflict detection slot")); ReplicationSlotCreate(CONFLICT_DETECTION_SLOT, false, RS_PERSISTENT, false, - false, false); + false, false, false); init_conflict_slot_xmin(); } diff --git a/src/backend/replication/logical/logical.c b/src/backend/replication/logical/logical.c index dc46a372ead..06efcaa60b2 100644 --- a/src/backend/replication/logical/logical.c +++ b/src/backend/replication/logical/logical.c @@ -108,9 +108,9 @@ static void LoadOutputPlugin(OutputPluginCallbacks *callbacks, const char *plugi * decoding. */ void -CheckLogicalDecodingRequirements(void) +CheckLogicalDecodingRequirements(bool repack) { - CheckSlotRequirements(); + CheckSlotRequirements(repack); /* * NB: Adding a new requirement likely means that RestoreSlotFromDisk() @@ -308,6 +308,7 @@ StartupDecodingContext(List *output_plugin_options, * output_plugin_options -- contains options passed to the output plugin * need_full_snapshot -- if true, must obtain a snapshot able to read all * tables; if false, one that can read only catalogs is acceptable. + * for_repack -- if true, we're going to be decoding for REPACK. * restart_lsn -- if given as invalid, it's this routine's responsibility to * mark WAL as reserved by setting a convenient restart_lsn for the slot. * Otherwise, we set for decoding to start from the given LSN without @@ -328,6 +329,7 @@ LogicalDecodingContext * CreateInitDecodingContext(const char *plugin, List *output_plugin_options, bool need_full_snapshot, + bool for_repack, XLogRecPtr restart_lsn, XLogReaderRoutine *xl_routine, LogicalOutputPluginWriterPrepareWrite prepare_write, @@ -344,7 +346,7 @@ CreateInitDecodingContext(const char *plugin, * On a standby, this check is also required while creating the slot. * Check the comments in the function. */ - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(for_repack); /* shorter lines... */ slot = MyReplicationSlot; diff --git a/src/backend/replication/logical/logicalfuncs.c b/src/backend/replication/logical/logicalfuncs.c index 9760818941d..512013b0ef0 100644 --- a/src/backend/replication/logical/logicalfuncs.c +++ b/src/backend/replication/logical/logicalfuncs.c @@ -115,7 +115,7 @@ pg_logical_slot_get_changes_guts(FunctionCallInfo fcinfo, bool confirm, bool bin CheckSlotPermissions(); - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(false); if (PG_ARGISNULL(0)) ereport(ERROR, diff --git a/src/backend/replication/logical/slotsync.c b/src/backend/replication/logical/slotsync.c index e75db69e3f6..28c89c5e10d 100644 --- a/src/backend/replication/logical/slotsync.c +++ b/src/backend/replication/logical/slotsync.c @@ -425,7 +425,7 @@ get_local_synced_slots(void) LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (int i = 0; i < max_replication_slots; i++) + for (int i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; @@ -814,6 +814,7 @@ synchronize_one_slot(RemoteSlot *remote_slot, Oid remote_dbid, */ ReplicationSlotCreate(remote_slot->name, true, RS_TEMPORARY, remote_slot->two_phase, + false, remote_slot->failover, true); @@ -1691,7 +1692,7 @@ update_synced_slots_inactive_since(void) LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (int i = 0; i < max_replication_slots; i++) + for (int i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; diff --git a/src/backend/replication/slot.c b/src/backend/replication/slot.c index 9533515a63b..47679161b67 100644 --- a/src/backend/replication/slot.c +++ b/src/backend/replication/slot.c @@ -151,6 +151,8 @@ ReplicationSlot *MyReplicationSlot = NULL; /* GUC variables */ int max_replication_slots = 10; /* the maximum number of replication * slots */ +int max_repack_replication_slots = 5; /* the maximum number of slots + * for REPACK */ /* * Invalidate replication slots that have remained idle longer than this @@ -188,14 +190,15 @@ static void SaveSlotToPath(ReplicationSlot *slot, const char *dir, int elevel); Size ReplicationSlotsShmemSize(void) { + int totalslots = max_replication_slots + max_repack_replication_slots; Size size = 0; - if (max_replication_slots == 0) + if (totalslots == 0) return size; size = offsetof(ReplicationSlotCtlData, replication_slots); size = add_size(size, - mul_size(max_replication_slots, sizeof(ReplicationSlot))); + mul_size(totalslots, sizeof(ReplicationSlot))); return size; } @@ -208,7 +211,7 @@ ReplicationSlotsShmemInit(void) { bool found; - if (max_replication_slots == 0) + if (max_replication_slots + max_repack_replication_slots == 0) return; ReplicationSlotCtl = (ReplicationSlotCtlData *) @@ -222,7 +225,7 @@ ReplicationSlotsShmemInit(void) /* First time through, so initialize */ MemSet(ReplicationSlotCtl, 0, ReplicationSlotsShmemSize()); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *slot = &ReplicationSlotCtl->replication_slots[i]; @@ -372,6 +375,7 @@ IsSlotForConflictCheck(const char *name) * db_specific: logical decoding is db specific; if the slot is going to * be used for that pass true, otherwise false. * two_phase: If enabled, allows decoding of prepared transactions. + * repack: If true, use a slot from the pool for REPACK. * failover: If enabled, allows the slot to be synced to standbys so * that logical replication can be resumed after failover. * synced: True if the slot is synchronized from the primary server. @@ -379,10 +383,11 @@ IsSlotForConflictCheck(const char *name) void ReplicationSlotCreate(const char *name, bool db_specific, ReplicationSlotPersistency persistency, - bool two_phase, bool failover, bool synced) + bool two_phase, bool repack, bool failover, bool synced) { ReplicationSlot *slot = NULL; - int i; + int startpoint, + endpoint; Assert(MyReplicationSlot == NULL); @@ -431,12 +436,16 @@ ReplicationSlotCreate(const char *name, bool db_specific, LWLockAcquire(ReplicationSlotAllocationLock, LW_EXCLUSIVE); /* - * Check for name collision, and identify an allocatable slot. We need to - * hold ReplicationSlotControlLock in shared mode for this, so that nobody - * else can change the in_use flags while we're looking at them. + * Check for name collision (across the whole array), and identify an + * allocatable slot (in the array slice specific to our current use case: + * either general, or REPACK only). We need to hold + * ReplicationSlotControlLock in shared mode for this, so that nobody else + * can change the in_use flags while we're looking at them. */ LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + startpoint = !repack ? 0 : max_replication_slots; + endpoint = max_replication_slots + (repack ? max_repack_replication_slots : 0); + for (int i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; @@ -444,7 +453,9 @@ ReplicationSlotCreate(const char *name, bool db_specific, ereport(ERROR, (errcode(ERRCODE_DUPLICATE_OBJECT), errmsg("replication slot \"%s\" already exists", name))); - if (!s->in_use && slot == NULL) + + if (i >= startpoint && i < endpoint && + !s->in_use && slot == NULL) slot = s; } LWLockRelease(ReplicationSlotControlLock); @@ -454,7 +465,8 @@ ReplicationSlotCreate(const char *name, bool db_specific, ereport(ERROR, (errcode(ERRCODE_CONFIGURATION_LIMIT_EXCEEDED), errmsg("all replication slots are in use"), - errhint("Free one or increase \"max_replication_slots\"."))); + errhint("Free one or increase \"%s\".", + repack ? "max_repack_replication_slots" : "max_replication_slots"))); /* * Since this slot is not in use, nobody should be looking at any part of @@ -547,7 +559,7 @@ SearchNamedReplicationSlot(const char *name, bool need_lock) if (need_lock) LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; @@ -575,7 +587,8 @@ int ReplicationSlotIndex(ReplicationSlot *slot) { Assert(slot >= ReplicationSlotCtl->replication_slots && - slot < ReplicationSlotCtl->replication_slots + max_replication_slots); + slot < ReplicationSlotCtl->replication_slots + + (max_replication_slots + max_repack_replication_slots)); return slot - ReplicationSlotCtl->replication_slots; } @@ -869,7 +882,7 @@ ReplicationSlotCleanup(bool synced_only) restart: found_valid_logicalslot = false; LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; @@ -1251,7 +1264,7 @@ ReplicationSlotsComputeRequiredXmin(bool already_locked) if (!already_locked) LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; TransactionId effective_xmin; @@ -1306,7 +1319,7 @@ ReplicationSlotsComputeRequiredLSN(void) Assert(ReplicationSlotCtl != NULL); LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; XLogRecPtr restart_lsn; @@ -1373,12 +1386,12 @@ ReplicationSlotsComputeLogicalRestartLSN(void) XLogRecPtr result = InvalidXLogRecPtr; int i; - if (max_replication_slots <= 0) + if (max_replication_slots + max_repack_replication_slots <= 0) return InvalidXLogRecPtr; LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s; XLogRecPtr restart_lsn; @@ -1453,11 +1466,11 @@ ReplicationSlotsCountDBSlots(Oid dboid, int *nslots, int *nactive) *nslots = *nactive = 0; - if (max_replication_slots <= 0) + if (max_replication_slots + max_repack_replication_slots <= 0) return false; LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s; @@ -1514,13 +1527,13 @@ ReplicationSlotsDropDBSlots(Oid dboid) bool found_valid_logicalslot; bool dropped = false; - if (max_replication_slots <= 0) + if (max_replication_slots + max_repack_replication_slots <= 0) return; restart: found_valid_logicalslot = false; LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s; char *slotname; @@ -1617,11 +1630,11 @@ CheckLogicalSlotExists(void) { bool found = false; - if (max_replication_slots <= 0) + if (max_replication_slots + max_repack_replication_slots <= 0) return false; LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (int i = 0; i < max_replication_slots; i++) + for (int i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s; bool invalidated; @@ -1655,17 +1668,24 @@ CheckLogicalSlotExists(void) * slots. */ void -CheckSlotRequirements(void) +CheckSlotRequirements(bool repack) { /* * NB: Adding a new requirement likely means that RestoreSlotFromDisk() * needs the same check. */ - if (max_replication_slots == 0) + if (!repack && max_replication_slots == 0) ereport(ERROR, - (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), - errmsg("replication slots can only be used if \"max_replication_slots\" > 0"))); + errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), + errmsg("replication slots can only be used if \"%s\" > 0", + "max_replication_slots")); + + if (repack && max_repack_replication_slots == 0) + ereport(ERROR, + errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), + errmsg("REPACK can only be used if \"%s\" > 0", + "max_repack_replication_slots")); if (wal_level < WAL_LEVEL_REPLICA) ereport(ERROR, @@ -2216,7 +2236,7 @@ InvalidateObsoleteReplicationSlots(uint32 possible_causes, Assert(!(possible_causes & RS_INVAL_WAL_REMOVED) || oldestSegno > 0); Assert(possible_causes != RS_INVAL_NONE); - if (max_replication_slots == 0) + if (max_replication_slots == 0 && max_repack_replication_slots == 0) return invalidated; XLogSegNoOffsetToRecPtr(oldestSegno, 0, wal_segment_size, oldestLSN); @@ -2224,7 +2244,7 @@ InvalidateObsoleteReplicationSlots(uint32 possible_causes, restart: found_valid_logicalslot = false; LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (int i = 0; i < max_replication_slots; i++) + for (int i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; bool released_lock = false; @@ -2329,7 +2349,7 @@ CheckPointReplicationSlots(bool is_shutdown) */ LWLockAcquire(ReplicationSlotAllocationLock, LW_SHARED); - for (i = 0; i < max_replication_slots; i++) + for (i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; char path[MAXPGPATH]; @@ -2430,7 +2450,7 @@ StartupReplicationSlots(void) FreeDir(replication_dir); /* currently no slots exist, we're done. */ - if (max_replication_slots <= 0) + if (max_replication_slots + max_repack_replication_slots <= 0) return; /* Now that we have recovered all the data, compute replication xmin */ diff --git a/src/backend/replication/slotfuncs.c b/src/backend/replication/slotfuncs.c index 9f5e4f998fe..16fbd383735 100644 --- a/src/backend/replication/slotfuncs.c +++ b/src/backend/replication/slotfuncs.c @@ -53,7 +53,7 @@ create_physical_replication_slot(char *name, bool immediately_reserve, /* acquire replication slot, this will check for conflicting names */ ReplicationSlotCreate(name, false, temporary ? RS_TEMPORARY : RS_PERSISTENT, false, - false, false); + false, false, false); if (immediately_reserve) { @@ -90,7 +90,7 @@ pg_create_physical_replication_slot(PG_FUNCTION_ARGS) CheckSlotPermissions(); - CheckSlotRequirements(); + CheckSlotRequirements(false); create_physical_replication_slot(NameStr(*name), immediately_reserve, @@ -146,7 +146,7 @@ create_logical_replication_slot(char *name, char *plugin, */ ReplicationSlotCreate(name, true, temporary ? RS_TEMPORARY : RS_EPHEMERAL, two_phase, - failover, false); + false, failover, false); /* * Ensure the logical decoding is enabled before initializing the logical @@ -164,6 +164,7 @@ create_logical_replication_slot(char *name, char *plugin, */ ctx = CreateInitDecodingContext(plugin, NIL, false, /* just catalogs is OK */ + false, /* not repack */ restart_lsn, XL_ROUTINE(.page_read = read_local_xlog_page, .segment_open = wal_segment_open, @@ -203,7 +204,7 @@ pg_create_logical_replication_slot(PG_FUNCTION_ARGS) CheckSlotPermissions(); - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(false); create_logical_replication_slot(NameStr(*name), NameStr(*plugin), @@ -240,7 +241,7 @@ pg_drop_replication_slot(PG_FUNCTION_ARGS) CheckSlotPermissions(); - CheckSlotRequirements(); + CheckSlotRequirements(false); ReplicationSlotDrop(NameStr(*name), true); @@ -270,7 +271,7 @@ pg_get_replication_slots(PG_FUNCTION_ARGS) currlsn = GetXLogWriteRecPtr(); LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); - for (slotno = 0; slotno < max_replication_slots; slotno++) + for (slotno = 0; slotno < max_replication_slots + max_repack_replication_slots; slotno++) { ReplicationSlot *slot = &ReplicationSlotCtl->replication_slots[slotno]; ReplicationSlot slot_contents; @@ -648,9 +649,9 @@ copy_replication_slot(FunctionCallInfo fcinfo, bool logical_slot) CheckSlotPermissions(); if (logical_slot) - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(false); else - CheckSlotRequirements(); + CheckSlotRequirements(false); LWLockAcquire(ReplicationSlotControlLock, LW_SHARED); @@ -665,7 +666,7 @@ copy_replication_slot(FunctionCallInfo fcinfo, bool logical_slot) * managed to create the new slot, we advance the new slot's restart_lsn * to the source slot's updated restart_lsn the second time we lock it. */ - for (int i = 0; i < max_replication_slots; i++) + for (int i = 0; i < max_replication_slots + max_repack_replication_slots; i++) { ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i]; diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c index 2bb3f34dc6d..9d7d675fa96 100644 --- a/src/backend/replication/walsender.c +++ b/src/backend/replication/walsender.c @@ -1220,7 +1220,7 @@ CreateReplicationSlot(CreateReplicationSlotCmd *cmd) { ReplicationSlotCreate(cmd->slotname, false, cmd->temporary ? RS_TEMPORARY : RS_PERSISTENT, - false, false, false); + false, false, false, false); if (reserve_wal) { @@ -1240,7 +1240,7 @@ CreateReplicationSlot(CreateReplicationSlotCmd *cmd) Assert(cmd->kind == REPLICATION_KIND_LOGICAL); - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(false); /* * Initially create persistent slot as ephemeral - that allows us to @@ -1251,7 +1251,7 @@ CreateReplicationSlot(CreateReplicationSlotCmd *cmd) */ ReplicationSlotCreate(cmd->slotname, true, cmd->temporary ? RS_TEMPORARY : RS_EPHEMERAL, - two_phase, failover, false); + two_phase, false, failover, false); /* * Do options check early so that we can bail before calling the @@ -1309,6 +1309,7 @@ CreateReplicationSlot(CreateReplicationSlotCmd *cmd) Assert(IsLogicalDecodingEnabled()); ctx = CreateInitDecodingContext(cmd->plugin, NIL, need_full_snapshot, + false, InvalidXLogRecPtr, XL_ROUTINE(.page_read = logical_read_xlog_page, .segment_open = WalSndSegmentOpen, @@ -1466,7 +1467,7 @@ StartLogicalReplication(StartReplicationCmd *cmd) QueryCompletion qc; /* make sure that our requirements are still fulfilled */ - CheckLogicalDecodingRequirements(); + CheckLogicalDecodingRequirements(false); Assert(!MyReplicationSlot); diff --git a/src/backend/utils/misc/guc_parameters.dat b/src/backend/utils/misc/guc_parameters.dat index a315c4ab8ab..bf82b8f6048 100644 --- a/src/backend/utils/misc/guc_parameters.dat +++ b/src/backend/utils/misc/guc_parameters.dat @@ -2071,6 +2071,14 @@ max => 'MAX_BACKENDS', }, +{ name => 'max_repack_replication_slots', type => 'int', context => 'PGC_POSTMASTER', group => 'REPLICATION_SENDING', + short_desc => 'Sets the maximum number of replication slots for use by REPACK.', + variable => 'max_repack_replication_slots', + boot_val => '5', + min => '0', + max => 'MAX_BACKENDS', +}, + /* see max_wal_senders */ { name => 'max_replication_slots', type => 'int', context => 'PGC_POSTMASTER', group => 'REPLICATION_SENDING', short_desc => 'Sets the maximum number of simultaneously defined replication slots.', diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 6d0337853e0..fb75990609e 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -348,6 +348,8 @@ # (change requires restart) #max_replication_slots = 10 # max number of replication slots # (change requires restart) +#max_repack_replication_slots = 5 # max number of replication slots for REPACK + # (change requires restart) #wal_keep_size = 0 # in megabytes; 0 disables #max_slot_wal_keep_size = -1 # in megabytes; -1 disables #idle_replication_slot_timeout = 0 # in seconds; 0 disables diff --git a/src/include/replication/logical.h b/src/include/replication/logical.h index bc9d4ece672..bc075b16741 100644 --- a/src/include/replication/logical.h +++ b/src/include/replication/logical.h @@ -115,11 +115,12 @@ typedef struct LogicalDecodingContext } LogicalDecodingContext; -extern void CheckLogicalDecodingRequirements(void); +extern void CheckLogicalDecodingRequirements(bool repack); extern LogicalDecodingContext *CreateInitDecodingContext(const char *plugin, List *output_plugin_options, bool need_full_snapshot, + bool for_repack, XLogRecPtr restart_lsn, XLogReaderRoutine *xl_routine, LogicalOutputPluginWriterPrepareWrite prepare_write, diff --git a/src/include/replication/slot.h b/src/include/replication/slot.h index 4b4709f6e2c..489af7d8d6c 100644 --- a/src/include/replication/slot.h +++ b/src/include/replication/slot.h @@ -324,6 +324,7 @@ extern PGDLLIMPORT ReplicationSlot *MyReplicationSlot; /* GUCs */ extern PGDLLIMPORT int max_replication_slots; +extern PGDLLIMPORT int max_repack_replication_slots; extern PGDLLIMPORT char *synchronized_standby_slots; extern PGDLLIMPORT int idle_replication_slot_timeout_secs; @@ -334,7 +335,7 @@ extern void ReplicationSlotsShmemInit(void); /* management of individual slots */ extern void ReplicationSlotCreate(const char *name, bool db_specific, ReplicationSlotPersistency persistency, - bool two_phase, bool failover, + bool two_phase, bool repack, bool failover, bool synced); extern void ReplicationSlotPersist(void); extern void ReplicationSlotDrop(const char *name, bool nowait); @@ -377,7 +378,7 @@ extern void ReplicationSlotDropAtPubNode(WalReceiverConn *wrconn, char *slotname extern void StartupReplicationSlots(void); extern void CheckPointReplicationSlots(bool is_shutdown); -extern void CheckSlotRequirements(void); +extern void CheckSlotRequirements(bool repack); extern void CheckSlotPermissions(void); extern ReplicationSlotInvalidationCause GetSlotInvalidationCause(const char *cause_name); -- 2.47.3 --cdhh4t7ukb6tnjoq-- ^ permalink raw reply [nested|flat] 354+ messages in thread
end of thread, other threads:[~2026-04-01 17:54 UTC | newest] Thread overview: 354+ messages (download: mbox mbox.gz follow: Atom feed) -- links below jump to the message on this page -- 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-03-24 14:09 [PATCH v44 07/10] Split cluster.h to create repack_internal.h Álvaro Herrera <alvherre@kurilemu.de> 2026-04-01 17:54 [PATCH v54 5/5] Reserve replication slots specifically for REPACK Álvaro Herrera <alvherre@kurilemu.de>
This inbox is served by agora; see mirroring instructions for how to clone and mirror all data and code used for this inbox